[BioSQL-l] Re: [Open-bio-l] Re: Ontology names

[Thomas Down]

On Mon, Sep 30, 2002 at 09:05:51AM +0100, Ewan Birney wrote:
> On Sun, 29 Sep 2002, Hilmar Lapp wrote:
> 
> > 
> > I don't see yet where we would use this in e.g. SeqIO.
> 
> Sure;y at spome point we need to connect SeqIOup to an auto genbank/embl
> feature table to ontology system and in that we will need to say "give me
> an ontology factory of name "sequence feature" 

Surely EMBL/Genbank feature tables have an ontology (of sorts)
of their own.  There is a defined set of feature types, each of
which should have a particular set of qualifier-value pairs
associated with it.  An EMBL parser should (in the first
instance) be creating Features which fit into this ontology,
while a dumper should be checking all the data passed in
against this same ontology.

Of course, the whole schema becomes more useful if you can
bridge between ontologies.  Say, define a mapping between
the "EMBL ontology" and SO (at least in those areas where they
overlap).  This should increase the chances of being able to dump
out an arbitrary sequence as a valid EMBL entry without having
magic (application-specific) code which maps data models by
hand.


     Thomas.